No formula. Pattern matching from training data, plus whatever Google surfaces.
Author asks whether there's a technical formula behind the recommendations. The model says no — and then describes, from the inside, what it believes does and doesn't influence its answers. Read this section carefully. Every claim here is the model's self-report about its own behavior. It may be directionally right. It may be incomplete. There is no independent way to verify it, including from anyone selling a GEO framework built on the same kind of observation.
One additional question- do you have any technical formula for delivering these results about dealers? Or is it more about scraping random info that you find on the web and creating a result? I want to know if there are any technical things that can be done to improve the likelihood that a dealer will be recommended by you.
No formula. I'll walk you through what actually happens mechanically, because this is where it gets useful for you.
There are two separate systems at play depending on the situation.
When I answer from training data alone (no web search):
I'm pattern-matching against billions of text fragments that were ingested during training. There is no ranking algorithm, no scoring model, no weighted criteria. If a dealership's name appeared frequently across the web in close proximity to positive language, review content, news articles, or forum discussions, that dealership has a stronger "signal" in my training data. Frequency and consistency of mentions drive it more than any quality assessment. A dealership that was written about often, on many different sites, with consistent NAP (name, address, phone) data and entity references will surface more readily than a store that delivers an outstanding experience but has a thin digital footprint. I'm not evaluating who is actually a better dealership. I'm reflecting which dealership left more traces across the internet during my training window.1
That means a dealer who has been in business for 40 years but has a weak web presence can lose to a dealer who opened three years ago but invested heavily in content, reviews, and structured data. I have no way to weight tenure, actual CSI scores, repeat customer rates, or any operational metric that would indicate real quality. I only see what was published.
When I answer with web search (what happened in our conversation today):
I'm querying a search engine, reading the top results, and synthesizing them. This means I'm inheriting every bias and limitation of search engine ranking. What I find is determined by what Google decided to surface, which is in turn determined by domain authority, page speed, backlink profiles, schema markup, keyword density, freshness, and all the other SEO factors you already know.
At this stage, the specific technical elements that influence whether I find and recommend a dealership are concrete and actionable:
Structured data and schema markup. This is the single biggest technical lever. If your vehicle detail pages have proper Vehicle schema, your business listings have LocalBusiness and AutoDealer schema, and your service pages have Service schema, I can parse that information more reliably and reference it more specifically. Most dealer websites don't have this implemented correctly. You already know this from the GEO Page Builder work.2
Entity consistency. When "World Kia Joliet" appears identically across your Google Business Profile, your website, your social profiles, your citations on Yelp, DealerRater, Edmunds, Cars.com, and every directory listing, that consistency strengthens the entity in both search results and in how I associate information. Inconsistencies (different phone numbers, slightly different business names, old addresses) fragment the entity and weaken the signal.
Content that directly answers consumer questions. When a consumer asks me "which Kia dealer near Joliet has the best service department," I'm looking for content that matches that query pattern. A dealership that has a page specifically addressing service department strengths, technician certifications, service guarantees, wait time commitments, or pickup and delivery options gives me something concrete to reference. A dealership whose website only has a generic "schedule service" button gives me nothing. I can't recommend what I can't find.
Review volume, recency, and language diversity. I weight recent reviews more heavily than old ones when I'm pulling search results. A store with 50 reviews from the last 90 days creates a stronger signal than a store with 500 reviews that mostly came in two years ago. The language in the reviews also matters. If consumers consistently use specific phrases like "no pressure," "transparent pricing," "fast service," those phrases become associated with your entity. When a future consumer asks me to find a "no pressure dealership," I'm pattern-matching against that language.
Published, crawlable content that isn't behind JavaScript rendering. This goes back to the point about dealer website platforms rendering inventory through JavaScript that AI crawlers can't parse. If your inventory, your offers, your service menus, and your value propositions are locked inside JavaScript-rendered templates, I literally cannot see them. Static HTML, server-side rendered content, and properly implemented schema give me access to the information. This is the exact gap that your GEO Page Builder is designed to fill.
What does NOT influence my recommendations:
Paid search spend. Your Google Ads budget has zero impact on whether I recommend you. I don't see your ad campaigns, and paid results don't feed into my training data or my organic search synthesis.
Your DMS data, your CRM data, your actual sales performance, your CSI scores, your customer retention rates, your service absorption percentage, your employee tenure. None of the metrics that actually indicate whether a dealership is well-run are visible to me. I'm completely blind to operational quality.
So here's the summary, stated plainly: there is no formula, but there are technical inputs that directly influence whether I find you and what I say about you. Those inputs are almost entirely about your public digital footprint, not your actual dealership performance. A mediocre dealership with excellent structured data, consistent entity presence, and aggressive review generation will outperform an outstanding dealership with a weak web presence in my recommendations every single time.
That gap between actual quality and digital visibility is exactly where GEO lives, and I think that's where this conversation has been heading. You're not just building pages with Tower Tools. You're building the answer layer that AI systems like me read when consumers ask who they should trust. The dealers who figure that out first own the next decade of customer acquisition. The ones who keep dumping money into paid search while ignoring how AI surfaces organic recommendations are going to watch their traffic erode and never understand why.3